Determinants of Cross Border Informal Trade: the

case of Benin

Sami Bensassi∗

Joachim Jarreau†

Cristina Mitaritonna‡

Preliminary Version

Abstract

This paper exploits a rich survey on informal cross border trade in Benin

(named ECENE), conducted by the National Institute of Statistics in 2011,

to study the determinants of informal versus formal trade. We concentrate on

export and re-exports of Benin to its neighboring countries. We �rst show that

goods traded formally and informally are very di�erent: the overlap between

the two types of trade is very thin, suggesting the existence of two separated

channels of trade, with agricultural products mainly traded informally. Second

we show that goods facing higher tari�s or an import ban explain, as expected,

the choice of informality. Finally, we �nd that trade policies are not the only

determinant of informality. Using an indicator of a product's time sensitiv-

ity, we �nd that this variable also contributes to explain the probability of

informal trade, in the case of agricultural products, suggesting that informal

trade facilitation through the payment of acceleration fees might also play an

important role.

Keywords: Smuggling, International trade, Trade policy, Time sensitiveness.

1 Introduction

Informal trade is pervasive in Africa, with important implications on the functioningof its economies. The evasion of customs duties is a serious concern for many Africancountries, for which tari�s account for a sizeable share of public receipts (Jeanand Mitaritonna, 2010). Moreover, informal cross border trade is also crucial tounderstand why, the level of intra-Africa trade is low by world standards. While theshare of in-trade reached 40% in North America and 63% in Western Europe sincemid-2000s, it was estimated at only 10 to 12% in Africa (UNECA, 2015).

On the empirical side, informal trade is inherently di�cult to measure. Fol-lowing the initial idea of Bhagwati (Bhagwati, 1964, 1967), several recent papershave analysed smuggling using mirror data that is, matching importer and exporter

∗University of Birmingham (sami.bensassi@bham.ac.uk).†University of Paris-Dauphine and IRD (joachim.jarreau@gmail.com).‡CEPII (cristina.mitaritonna@cepii.fr).

1

reported bilateral trade �ows. If the source country's reported exports exceed thedestination country's reported imports, smuggling can be inferred (Fisman and Wei,2004; Javorcik and Narciso, 2008; Mishra et al., 2008; Jean and Mitaritonna, 2010;Bouet and Roy, 2012).

Intra-African smuggling, however, cannot be fully apprehended with mirror data,since o�cial trade statistics from both the exporting and importing country failto record cross-border trade. For instance the recorded trade between Benin andNigeria is very small.

In this study we exploit a rich survey on informal cross border trade in Benin(named ECENE), conducted by the National Institute of Statistics in 2011, to studythe determinants of informal trade compared to formal one. We concentrate onexport and re-exports of Benin its neighboring countries. As main source of formaltrade we use the COMTRADE database, considering the declaration of Benin versusthe same destinations in 2011.1

Macroeconomic researches trying to determine the level of informal trade and itsdeterminant(Ayadi et al., 2013; Golub and Mbaye, 2009; Golub, 2012) identify thegood smuggled, gather data on their prices across borders and checked whether theprice di�erence observed come from the tari�s, taxes or ban applied in the countriestrading informally.

In this paper we also look as main determinants at trade policies variables, suchas tari�s applied by importers or bans imposed by Nigeria, and time sensitiveness.

We �rst show that the two forms of trade are very di�erent: the overlap betweengoods appearing in each of the databases is very thin, suggesting the existence oftwo separated channels of trade, with agricultural products mainly traded informally.Second we show that goods facing higher tari�s explain, as expected, the the choiceof informality. However tari�s do not play anymore a relevant role in goods facingan import ban in Nigeria, in this case the ban becomes the main signi�cant variable.Finally, we �nd that trade policies are not the only determinant of informality. Usingan indicator of a product's time sensitivity, we �nd that this variable also contributesto explain the probability of informal trade, in the case of agricultural products,suggesting that informal trade facilitation through the payment of acceleration feesmight also play an important role.

The remaining of this paper is organized as follows: in section 2 we remind thecontext in which the informal trade relations have grown in Benin. In section 3 wedescribe the informal and formal Benin's bilateral trade patterns. In section 4 wepresent the empirical strategy used to look at the main determinants of informaltrade �ows and we present the results. Section 6 concludes.

2 Entrepot trade in Benin: context

Benin is a French-speaking country in West Africa with a population of 10 million.It started to be an Entrepot economy at the end of the 1960's (Igue and Soule, 1992).

1It has to be noticed that the COMTRADE UN database is not deprived of problems. The qual-ity of the data depends on the reliability of the reports made by custom agencies. In addition valuereported can be subjected to under invoicing and the nature of the products to mis-classi�cation

2

Lacking natural resources and an industrial basis, but bene�ting from a good accessto the Atlantic Ocean through the port of Cotonou, the country positioned itselfas a hub for trade �ows to its landlocked neighboring countries, Burkina Faso andNiger, as well as to Nigeria, its giant neighbor country (170 million inhabitants).

In the case of Nigeria, rather than the access to the ocean, other factors havecontributed to make trade with Benin attractive. Among these were the politicalturmoil associated with the Biafra war between 1967 and 1970, the increase of oilrevenues in Nigeria after 1974, and, importantly, the protectionist trade policy inthis country. Nigeria adopted an import substitution strategy of industrializationsoon after its independence, with a concomitant high level of tari�s, as well as alist of import prohibitions for various goods. Despite the country's adoption ofa structural adjustment program in 1986 (in theory implying a more open tradepolicy), the participation to the World Trade Organization from its start in 1995,the participation to the Economic Community of West African States (ECOWAS),and the adoption of ECOWAS common external tari�s in 2005, a number of importbans are still in place to this date (Volker, 2010). By contrast, Benin adopted anopen trade policy after 1989. Therefore, trade protection levels in the two countriesdi�er widely. Numerous consumption goods, such as edible oil or poultry meat, arebanned for import in Nigeria, while facing no or low tari�s in Benin.

This situation has led to the development of trade between Benin and Nigeria ona large scale. This trade involves imports from third countries, which are importedinto Benin at the port of Cotonou, before crossing the land border to Nigeria, aswell as products of Benin, or of other West African countries. A large part of thistrade is informal, i.e. illegal and unrecorded (Raballand and Mjekiqi, 2010).

As the map in �gure 1 shows, the port of Cotonou is close to Lagos and tothe south-western region of Nigeria, where a large part of economic activity con-centrates. Qualitative accounts show how goods are transported through variousroutes, crossing the Benin-Nigeria border at several points along the border. Animportant part of the trade also transits through Niger, to reach the large marketof Kano in north-west Nigeria (Walther, 2015).

Having goods transit through Cotonou may involve only a small increase indistance, and this may be compensated by gains in cost and time associated withthis route. Similarly, exporting goods from Benin to Nigeria may involve a choicebetween a formal, or legal route, and an informal one. Di�erent protection levelsclearly create incentives for smuggling. In addition, case studies on the case of Benin(Bako-Arifari, 2001; ?) suggest that weak enforcement at government agency levels,as well as the slowness of custom procedures, could also play a role. In the followingsections, we study some of these determinants of the choice of informal trade, usingdata for both formal and informal trade �ows.

3 Data and descriptive statistics

3.1 Data

In order to study the determinants of informal trade compared to formal one inBenin, we use di�erent sources of data. For informal (not recorded) trade we use

3

Figure 1: Map of Benin and its neighbours

the ECENE survey, which has been conducted by the national institute of statisticsof Benin in 2011. Its objective has been to quantify informal trade at Benin borders.To realize this, 171 border crossing passages identi�ed as actively used by smugglerswere surveyed in the month of September 2011. Information through questionnaireaddressed to informal traders were gathered on the nature, quantity and the valueof smuggled goods. The ECENE survey contains information at a very detail levelof products (HS10) crossing the Benin borders in di�erent ways: export, re-exports,import and transit. 8883 border crossings were recorded and 10415 �ow of goodsidenti�ed (INSAE, 2011). In this �rst part of our research we concentrate on exportand re-exports of Benin to Togo, Niger and Nigeria, the three destinations with thelargest frequency of informal trade exchanges. We are aware that the largest part oftrade activities of Benin is done by transit however at the moment we do not have anyinformation on the transit activity registered by customs in Benin in 2011. Formaltrade (or registered trade) is recorded in the COMTRADE UN trade database. Datais available in both databases for the year 2011. However, COMTRADE contains�ows only at the country level using the HS6 Rev3 nomenclature. For the sake ofcomparison, the full ECENE database has been aggregated accordingly.

For trade policies measures, we use the MFN tari�s as well as the preferentialtari�s by HS6 product applied by the di�erent Benin's neighboring countries for theyear 2011. Tari�s have been provided by the customs of each country. An importantwork has been done to reconstruct a complete database of Nigerian Bans at theproduct level for the year 2011, using documents published by Nigeria's minister ofFinance and WTO reports. At the end of the nineties, the government of Nigeriahad a plan to phase out import bans. 23 products were facing bans in 1998, and anumber of these were replaced by high tari�s between 1999 and 2001. However, thetrend reversed in 2002(WTO, 2005). 218 products (HS4) had an import ban at theend of 2004. This prohibition list remained essentially unchanged until the end of2008, when a number of products were removed from the list. This was in part a

4

consequence of Nigeria's membership of ECOWAS, which required it to align withthe group's common external tari�. However, numerous products remained on theprohibition list after 2008, in a context where the implementation of the ECOWAScommon tari� was subject to negotiations between Nigeria and other members, andwas repeatedly delayed.

Finally the information for time sensitiveness at the HS4 level have been kindlyprovided by Hummels and Schaur Hummels and Schaur (2013).

3.2 Descriptive statistics by product in informal and formal

trade

Table 1 provides some statistics by product recorded in the ECENE database, forre-export and export of Benin with its neighboring countries.

Table 1: ECENE Database - re-export and export by product (2011)re-export: Nbr of products export: Nbr of products

Agregation level HS6 HS4 HS2 HS6 HS4 HS2DestinationGhana 3 (0) 2 2 0 (0) 0 0Niger 7 (1) 7 6 10 (9) 10 6Nigeria 73 (26) 55 33 112 (59) 86 38Togo 26 (4) 22 19 110 (48) 88 43Burkina 1 (0) 1 1 2 (1) 2 2Total 110 (31) 87 61 234 (117) 186 89In parentheses the number of agricultural products, as de�ned by the WTO, excl. processed food.

Several points are worth highlighting. Firstly, the main countries to which Benintrade informally with are Nigeria, Togo and Niger. In the following we will focus onthese three destinations. Secondly, from column 1 and column 4, we can notice thatthe distribution of products is balanced between agricultural and non-agriculturalones, in particular in the case of export. Finally, if we look at the number of productsby the di�erent aggregation level (HS6, HS4 or HS2), we can say that products arequite widespread across sectors.

The following relevant aspects emerge when comparing informal to formal trade,by destination (see table 2):

• the number of informal products are quite important, often more than formalones. In particular agricultural products are more frequent in the ECENEdatabase.

• the two types of trade seem to be complementary. Indeed, very few of the HS6products are recorded in both databases, providing evidence of two separatedchannels of trade. For re-export there is only one product at the HS6 level,that is re-exported informally and formally to Togo (code 271019, which is oil)and also only one at the HS4 level (code 2710); while in the case of export we�nd 5 HS6 products exported in both ways to Nigeria and 11 to Togo, mainlyindustrial items.

5

Table 2: ECENE and COMTRADE by product (2011)re-export export

Nb HS6 Nb HS6 HS6 Nb HS6 Nb HS6 HS6Destination ECENE COMTRADE Common pdts ECENE COMTRADE Common pdtsNiger 7 (1) 5 (0) 0 10 (9) 41 (3) 0Nigeria 73 (26) 15 (0) 0 112 (59) 47 (5) 5(2)Togo 26 (4) 19 (1) 1 (0) 110 (48) 89 (7) 11 (2)In parentheses the number of agricultural products, as de�ned by the WTO, excl. processed food.

Table 3: ECENE and COMTRADE by product under a ban in Nigeria (2011)re-export export

Nb HS6 with a Ban2011 Nb HS6 with a Ban2011 Nb HS6 with a Ban2011 Nb HS6 with a Ban2011Destination ECENE COMTRADE ECENE COMTRADENiger 1 (0) 0 3 (2) 13(0)Nigeria 28 (8) 2 (0) 29(13) 12(4)Togo 0 0 0 0In parentheses the number of agricultural products, as de�ned by the WTO, excl. processed food.

Note: we consider that the ban imposed in Nigeria is not relevant in the case of Togo.

6

4 Regressions results. Trade policies as main de-

terminants

In this part we measure the importance of trade policies measures as main determ-inants of informal cross-border trade, compared to formal one. We �rst perform alinear probability model, then as robustness check a probit model to test the impactof tari� and bans for a goods to appear in informal, rather than in formal trade.We treat separately the case of re-export and export, in subsection 4.1 and in sub-section 4.2, respectively. Results we obtain at this stage are very similar for bothcases. In both database we drop double observations at the HS6 level for any givendestination. Moreover, as destinations we only consider Niger, Nigeria and Togo,the three destinations with the largest frequency of informal trade exchanges.

The dependent variable takes the value of 1 if the HS6 product is recorded asinformal and 0 otherwise. As tari� we take the MFN in the case of re-export andthe preferential tari� in the case of exports, that importers apply at the HS6 level.

Note that we de�ne the Ban2011 variable as an indicator being 1 for HS6products for which a ban has been implemented in Nigeria in 2011 at least to oneproduct within the HS6 category. We just leave the same value also for Niger, as itis well known that often goods under a ban in Nigeria are smuggled through Niger.

4.1 Results for re-export

In table 4 we use a linear probability model to test the impact of import bans andtari�s on the probability for a goods to appear in informal, rather than in formaltrade, for re-export.

It is important to notice that in each regressions we control for destination �xede�ects, and we introduce a dummy for agricultural products. The inclusion of such adummy is dictated primarily by the fact that agricultural products in re-exports areexclusively registered in the ECENE database. From column 1 we can see that thee�ect of tari� are positive and signi�cant in explaining the informality. The same incolumn 2 for Ban2011 alone. In column 3 the e�ect of Ban2011 disappears, becausea collinearity problem. Indeed a product facing a ban is often subject to a low tari�,but in the case where there is a ban applied, tari� is not important anymore, asthe ban is a prohibitive barrier. We introduce an interaction in column 4, whichin fact con�rms this;2 the variable Ban2011 returns to be positive and statisticallysigni�cant. From columns 5 to 7 we show results at country level. We can noticethat tari� play a crucial role in the case of Togo, while in the case of Nigeria thevariable ban is the main determinant, as well as when considering Niger and Nigeria.

When we perform a probit model results from the linear model still hold (seetable 5).

2Indeed the di�erence between the coe�cient for tari� plus standard error and the coe�cientof interaction minus standard error overlap

7

Table 4: Determinants of informal and formal trade: re-export. Linear model.Dep. variable:1 if Ecene , 0 if Comtrade

(1) (2) (3) (4) (5-Tgo) (6-Nga) (7- Ner and Nga)MFN2011 2.618*** 2.525*** 3.424*** 5.450*** 2.474*** 2.520***

(0.455) (0.467) (0.539) (0.846) (0.677) (0.613)Di� MFN -2.545 -2.570 -2.979 -2.859

(2.100) (2.095) (2.108) (2.135)Ban2011 0.182** 0.062 0.791*** 0.658*** 0.664***

(0.071) (0.079) (0.164) (0.193) (0.182)MFN2011*Ban2011 -4.241*** -3.274*** -3.316***

(0.752) (0.930) (0.852)Di� MFN*Ban2011 5.150 5.022

(2.556) (2.543)Agr Dummy 0.210*** 0.261*** 0.213*** 0.296*** 0.317** 0.291***

(0.058) (0.060) (0.060) (0.059) (0.135) (0.060)Destination F..E Yes Yes Yes Yes YesN 144 144 144 144 44 88 100R2 0.304 0.158 0.306 0.376 0.446 0.281 0.322Standard errors in parentheses. All regressions are clustered at the HS6 level.

Each speci�cation includes a constant term.

Di� MFN measures by product the di�erence in tari� between Nigeria and Niger.The variable is always zero for Togo an Nigeria

* p<0.1, ** p<0.05, *** p<0.001

Table 5: Determinants of informal and formal trade: re-export. Probit Model.Dep. variable:1 if Ecene , 0 if Comtrade

(1) (2-Tgo) (3-Nga) (4- Ner and Nga)MFN2011 13.138*** 17.754*** 10.411*** 10.110***

(2.779)) (4.152) (3.611) (3.123)Di� MFN -9.491 -8.691

(6.736) (6.539)Ban2011 2.313*** 2.075** 2.040**

(0.900) (0.960) (0.932)MFN2011*Ban2011 -12.395** -9.795* -9.495*

(5.200) (5.777) (5.482)Di� MFN*Ban2011 (empty) (empty)

Agr Dummy 1.743*** 1.297**(0.461) (0.606)

Destination F..E Yes YesN 143 44 62 72PseudoR2 0.374 0.376 0.235 0.250Standard errors in parentheses. All regressions are clustered at the HS6 level.

Each speci�cation includes a constant term.

Di� MFN measures by product the di�erence in tari� between Nigeria and Niger.The variable is always zero for Togo an Nigeria

* p<0.1, ** p<0.05, *** p<0.001

8

4.2 Results for export

For sake of comparison with he re-export case, we �rst perform in column 1-column 4regressions including an agriculture dummy. Then we also try regressions includingHS2 �xed e�ects in column 5-column 8 see table 6.

The results are similar even if the magnitude of the results including product�xed e�ects is more attenuated than those with the dummy agriculture.

Table 6: Determinants of informal and formal trade: export. Linear model.Dep. variable:1 if Ecene , 0 if Comtrade

Agr Dummy HS2 FE(1) (2 - Tgo) (3 Nga) (4 -Ner and Nga ) (5) (6-Tgo) (7-Nga) (8- Ner and Nga)

Bil Tari� 1.198*** 1.684*** 1.462** 0.793* 0.710* 1.081* 0.211 0.455(0.417) (0.613) (0.584) (0.476) (0.362) (2.165) (0.482) (0.005)

Di� MFN Bil 2.156 2.554*** 1.125** 2.165**(0.428) (0.459) (0.501) (2.165)

Ban2011 0.471** 0.534*** 0.454** 0.253* -0.074 -0.111(0.194) (0.204) (0.198) (0.175) (0.235) (0.242)

Bil*Ban2011 -2.809*** -3.263*** -2.586** -1.607*** -0.318 -0.133(1.236)) (1.262) (1.358) (0.934) (1.710) (1.605)

Di� MFN Bil*Ban2011 omitted omitted

Agr Dummy 0.417*** 0.374*** 0.389*** 0.465***(0.047) (0.070) (0.056) (0.057)

Destination F..E Yes Yes Yes Yes Yes YesHS2 F..E Yes Yes Yes YesN 409 199 159 210 409 199 159 210R2 0.318 0.247 0.211 0.394 0.603 0.579 0.699 0.716Standard errors in parentheses. All regressions are clustered at the HS6 level.

Each speci�cation includes a constant term.

We apply to Niger the same structure of protection in Nigeria

* p<0.1, ** p<0.05, *** p<0.001

Table 7: Determinants of informal and formal trade: export. Probit model.Dep. variable:1 if Ecene , 0 if Comtrade

Agr Dummy HS2 FE(1) (2 - Tgo) (3 Nga) (4 -Ner and Nga ) (5) (6-Tgo) (7-Nga) (8- Ner and Nga)

Bil Tari� 4.411*** 5.421*** 5.005** 3.146* 3.603* 5.074* 6.516 2.565(1.392) (1.754) (2.113) (1.778) (1.892) (2.752) (5.400) (3.358)

Di� MFN Bil 7.167*** 8.191*** 5.031** 12.994***(1.500) (1.554) (0.261) (4.012)

Ban2011 1.996** 2.263** 2.217** 1.556* -0.581 -0.973(0.833) (0.941) (0.967) (0.960) (1.682) (1.683)

Bil*Ban2011 -12.029*** -14.121*** -12.867** -9.243* -6.632 -0.520(6.254)) (7.772) (7.966) (5.129) (10.739) (9.970)

Di� MFN Bil*Ban2011 omitted omitted

Agr Dummy 0.417*** 1.267*** 1.645*** 1.930***(0.047) (0.260) (0.301) (0.315)

Destination F..E Yes Yes Yes Yes Yes YesHS2 F..E Yes Yes Yes YesN 409 199 159 210 299 133 55 89R2 0.280 0.210 0.211 0.362 0.420 0.318 0.159 0.374Standard errors in parentheses. All regressions are clustered at the HS6 level.

Each speci�cation includes a constant term.

We apply to Niger the same structure of protection in Nigeria

* p<0.1, ** p<0.05, *** p<0.001

9

5 Regressions results. Looking for other determin-

ants: time sensitivity

In this section we try to test whether informal trade facilitation through the paymentof acceleration fees might also to have an importance. As a measure of time sensit-iveness we use estimation provided by Hummels and Schaur. The authors in theirpaper (Hummels and Schaur, 2013) estimated a model exporters' choice betweenfast, expensive air cargo and slow, cheap ocean cargo, which depends on the priceelasticity of demand and the value that consumers attach to fast delivery. Usingimports data for the US they provide rich variation in the premium paid for airshipping and in time lags for ocean transit to extract consumers' valuation of time.They estimate that each day in transit is equivalent to an ad-valorem tari� of 0.6 to2.1 percent. Authors �nd a di�erent e�ects for the two categories of products: fresh(perishable goods) and part and components. For perishable goods they �nd thata higher "fresh" share increases the use of air shipment, but does not signi�cantlyinteract with transit days, suggesting that products such as "fresh �sh" are so timesensitive that any delay longer than a few days ruins the product. The e�ect showsup entirely in a higher use of air shipment for all exporters, regardless of oceantransit time to the US. For parts and components instead and components shareof trade for a given exporter-HS6 product results in a sharp increase in the timesensitivity of that trade. Comparing a product with zero component share to onethat is 100 percent components raises time sensitivity by 60 percent.

Hummels and Schaur provided us with unpublished estimations of time sensit-iveness at the HS4 level. Unfortunately information are not always available for allthe products in our database. In particular in the case of re-export we can test thisdeterminant only for industrial products. Moreover the distribution of the indicatoris very noisy, with some negative values. This is likely for two reasons, as suggestedby authors. First, there is an issue of sampling or there are some products thathave very low observation counts with very imprecise estimating values. As we donot have information on the number of observations by products or the standarddeviation of the index of time sensitiveness estimated, at this stage to deal withnegative values we normalize the distribution of the index, subtracting from eachobservation the minimum value and dividing by the standard deviation. Moreoverwe drop the most extreme values.

In table 8 we show linear estimations for re-export (column 1 and 2) and export(column 5-10), including this measure of time sensitiveness. As we can see in column1, column 3 and column 6, estimations for trade policies variables do not changesigni�cantly when using only the sample of goods for which we have information ontime sensitiveness.

The measure of time sensitiveness does not seem to be an important determinantof informal trade, when considering industrial products, while becomes positive whenincluding agricultural products.

10

Table 8: Adding time sensitiveness. Linear model.Dep. variable:1 if ECENE , 0 if COMTRADE

re-export export(1 Industry) (2 Industry) (1 All) (2 All) (3 Industry) (1 All) (2 All) (3 Industry)

MFN2011 3.546*** 3.822*** Bil Tari� 1.420*** 1.452*** 1.720*** 1.050*** 1.195*** 1.390***(0.659) (0.661) (0.463) (0.452) (0.530) (0.402) (0.413) (0.467)

Di� MFN -2.810 -1.999 Di� MFN Bil 2.079*** 2.117*** 2.498*** 1.708*** 1.941*** 2.214***(2.287) (2.338) (0.502) (0.485) (0.641) (0.610) (0.627) (0.776)

Ban2011 0.692*** 0.748*** Ban2011 0.607** 0.617** 0.569* 0.263 0.276 0.269(0.186) (0.186) (0.305) (0.298) (0.452) (0.256) (0.259) (0.285)

MFN2011*Ban2011 -3.458*** -3.740*** Bil*Ban2011 -3.660** -3.645** -3.087* -1.773 -1.719 -1.452(0.956) (0.961) (1.870) (1.809) (2.616) (1.432) (1.421) (1.831)

Di� MFN*Ban2011 5.519 4.606 Di� MFN Bil*Ban2011 omitted omitted omitted omitted omitted omitted

Time sensitiveness 0.685 Time sensitiveness 0.088*** 0.050 0.131* 0.098(0.704) (0.030) (0.88) (0.088) (0.992)

HS2 F.E. No No HS2 F.E. No No No Yes Yes YesDest. F.E. Yes Yes Dest. F.E. Yes Yes Yes Yes Yes YesN 104 103 N 324 324 170 324 324 249R2 0.312 0.332 R2 0.287 0.300 0.406 0.587 0.587 0.547Standard errors in parentheses. All regressions are clustered at the HS6 level.

Each speci�cation includes a constant term.

* p<0.1, ** p<0.05, *** p<0.001

6 Conclusion

In this paper, we study the determinants of informal cross-border trade betweenBenin and its neighbors. We combine two exhaustive trade data sources, the ECENEdatabase, for goods traded informally and the COMTRADE database, for goodsexchanged formally. The ECENE database is a survey of informal transactionsconducted by INSAE in Benin in 2011; the COMTRADE database gathers all formalinternational trade �ows. We compare goods appearing in formal and informaltransactions in 2011; we focus for now on export and re-export �ows between Beninand its neighbors (mainly Togo, Nigeria, and Niger). As main determinants we lookat trade policies variables, such as tari�s applied by importers or bans imposed byNigeria.

We �rst show that these goods are very di�erent: the overlap between goodsappearing in each of the databases is very thin, suggesting the existence of twoseparated channels of trade. Second we show that goods facing higher tari�s explain,as expected, the the choice of informality. However tari�s do not play anymore arelevant role in goods facing an import ban in Nigeria, in this case the importban becomes the main signi�cant variable. Finally, we �nd that trade policies arenot the only determinant of informality. Using an indicator of a product's timesensitivity, we �nd that this variable also contributes to explain the probability ofinformal trade, in the case of agricultural products, exported in majority throughthe informal channel, in particular concerning the agricultural goods, suggestingthat informal trade facilitation through the payment of acceleration fees might alsohave an important role.

References

Ayadi, L., N. Benjamin, S. Bensassi, and G. Raballand (2013). Estimating informaltrade across tunisia?s land borders. Policy Research Working Paper 6731, TheWorld Bank, Washington, DC.

11

Bako-Arifari, N. (2001). La corruption au port de cotonou : douaniers et intermé-diaires. Politique africaine 3 (83), 38�58.

Bhagwati, J. (1964). On the underinvoicing of imports. Oxford Bulletin of Economics

and Statisitcs 26 (4), 389�97.

Bhagwati, J. (1967). Fiscal Policies, the Faking of Foreign Trade Declarations, andthe Balance of Payments. Oxford Bulletin of Economics and Statisitcs 30, 61�67.

Bouet, A. and D. Roy (2012). Trade protection and tax evasion: Evidence fromkenya, mauritius, and nigeria. The Journal of International Trade and Economic

Development 21 (2), 287�320.

Fisman, R. and S. Wei (2004). Tax Rates and Tax Evasion: Evidence from ?MissingImports? in China'. Journal of Political Economy 112 (2), 471�496.

Golub, S. S. (2012). Entrepot trade and smuggling in west africa: Benin, togo andnigeria. The World Economy 35 (9), 1139�1161.

Golub, S. S. and A. A. Mbaye (2009). National trade policies and smuggling inafrica: The case of the gambia and senegal. World Development 37 (3), 595 � 606.

Hummels, D. L. and G. Schaur (2013). Time as a Trade Barrier. American Economic

Review 103 (7), 2935�59.

Igue, J. O. and B. Soule (1992). L'Etat Entrepot au Benin: Commerce Informal ou

Reponse a la Crise? Khartala: Paris.

INSAE (2011). Enquête sur le commerce non enregistré au cordon douanier. Tech-nical report, Institut National de la Statistique et de l'Analyse Economique,Cotonou.

Javorcik, B. S. and G. Narciso (2008). Di�erentiated products and evasion of importtari�s. Journal of International Economics 76 (2), 208�222.

Jean, S. and C. Mitaritonna (2010). Determinants and pervasiveness of the evasionof custom duties. Cepii Working Paper 2010-26, Centre d'etudes prospectives etd'information internationales.

Jibao, S., W. Prichard, and V. van den Boogard (2015). Pro�t, networks, powerand control: The multiple facets of informal cross-border trade practices in sierraleone. Unpublished.

Mishra, P., A. Subramanian, and P. Topalova (2008). Tari�s, enforcement, andcustoms evasion: Evidence from india. Journal of public Economics 92 (10), 1907�1925.

Raballand, G. and E. Mjekiqi (2010). Nigeria�s Trade Policy Facilitates Unof-�cial Trade but not Manufacturing. In Putting Nigeria to Work: A Strategy for

Employment and Growth (World Bank ed.). Washington D.C.: Volker Treichel.

12

Titeca, K. (2012). Tycoons and contraband: Informal cross-border trade in north-western uganda. Journal of Eastern African Studies 6 (1), 47�63.

UNECA (2015). Industrilizing through trade. Economic report on africa, UNECA.

Volker, T. (2010). Employment and growth in nigeria. In V. Treichel (Ed.), PuttingNigeria to Work: A Strategy for Employment and Growth, Chapter 1. Washington,D.C: World Bank.

Walther, O. J. (2015). Business, brokers and borders: The structure of west africantrade networks. The Journal of Development Studies 51 (5), 603�620.

WTO (2005). Trade policy review - nigeria. Technical report, World Trade Organ-ization, Geneva.

13